AFECL

Edge Contrastive Learning: An Augmentation-Free Graph Contrastive Learning Model

Requirements

This code package was developed and tested with Python 3.9

The required dependencies are as follows:

• numpy 1.26.4
• torch 1.12.0
• dgl 1.1.1

How to run

To run AFECL：

python train.py --dataset=cora --lr=0.01 --num-heads=4 --num-hidden=32 --tau=1 --weight-decay=1e-4 --in-drop=0.6 --attn-drop=0.5 --epochs=2000 --num-layers=1 --seed=1 --negative-slope=0.2 --rate=1

Citation

@inproceedings{li2025edge,
  title={Edge Contrastive Learning: An Augmentation-Free Graph Contrastive Learning Model},
  author={Li, Yujun and Zhang, Hongyuan and Yuan, Yuan},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  volume={39},
  number={17},
  pages={18575--18583},
  year={2025}
}